Design And Implementation Of Human Detection And Tracking System Based On Deep Learning

With the development of computer technology,people's requirements for artificial intelligence are increasing.Currently,it is a research hotspot to solve the problem of human tracking with using computer vision.Moving object detection and tracking technology provide strong technical support for the arrival of video monitoring automation,especially in intelligent video surveillance.Moving object detection and tracking technology are the basic technologies and two main research directions of computer vision technology.In this paper,the fusion method of three-frame difference and Gaussian mixture model is used to realize the detection of moving object detection in video,the method of deep learning is used to realize the recognition and tracking of human body.The main work is as follows.1)The video sequence meeting the requirements of target tracking is obtained and preprocessed.After obtaining the video data,filter algorithm and mathematical morphology operation are used to preprocess the original video data to make the image information more complete and more precise,and it is convenient for moving human objects to detect and track in the following research.2)The moving objects are detected.First of all,the three-frame difference algorithm is studied and implemented,which has a good detection result under the condition of poor illumination.Then,the moving object detection based on Gaussian mixture model is achieved,which effectively avoids the content hole problem of difference between three frames.But the detection is sensitive to illumination conditions.The fusion of three frame difference and mixed Gaussian model algorithm is used to detect moving targets.It avoids the double shadow phenomenon caused by ordinary two-frame difference.At the same time,it also solves the problems of large holes in the detected target caused by three-frame difference and the problems of the initial modeling of mixed Gaussian mode target detection and the detection depending on good illumination.Experimental results demonstrate that better detection effect is achieved.3)The human target is tracked.The method of deep learning is used to track the moving object.The principle of DeepSORT is analyzed.In the target tracking of DeepSORT,the moving target detection method proposed in this paper is taken as a part of DeepSORT,which can effectively solve the problem of tracking instability in complexillumination environment.At the same time,VGGNet is used to distinguish pedestrians and other moving objects in the tracking process,and it exerts an expected tracking influence.According to the above research,the human detection and tracking system based on deep learning is introduced.The system can detect and track the moving human objects in the video in real time,which can meet the needs of users.Finally,the test shows that the system achieves the expected design goals.